461

Periventricular White-Matter Cysts in a Murine Model of Gram-Negative Ventriculitis Thomas P. Naidich,' David G. McLone,2 and Yasuo Yamanouchi 2

Hydrocephalic patients with shunt infections frequently de­velop multiple cerebrospinal-fluid-density cysts that cause mid­line shift and life-threatening intracranial hypertension and re­spond poorly, if at all, to shunt diversion of cerebrospinal fluid. These cysts have been considered to represent multiloculation of the ventricular system by ependymal adhesions and veils resulting from ventriculitis. Studies using an experimental model of E. coli meningitis/ventriculitis in the hy-3 mouse suggest these cysts: (1) develop by the coalescence of lakes of white­matter edema, (2) grow to large size entirely within the periven­tricular white matter, and (3) cause pseudoloculation of the ventricle by compression from without. The so-called intraven­tricular septa or "veils" are the ependyma displaced inward by subependymal cysts or sheets of residual pericystic white mat­ter. This finding permits better interpretation of computed to­mographic images depicting persistent enlargement of the so­called multiloculations despite functioning ventricular shunt catheters, the multiplicity of cysts, and the white-matter location of these cysts.

Hydrocephalic children with shunt infection , choroiditis, and ven­triculitis frequently suffer: (1) reduced cerebrospinal fluid (CSF) production, (2) poor ventricular CSF turnover, (3) elevated CSF protein concentration, (4) isolation of individual ventricles by para­ventricular abscesses or cysts that compress the exit foramina, and (5) so-called multiloculation of the ventricular system by broad sheets of tissue designated " intraventricular septations," "septa, " or " ventricular veils" [1-10]. Such children frequently require or­ganism-specific antibiotics, immediate placement of external ven­tricular drains, multiple revisions of obstructed shunts, placement of long multihole catheters to attempt drainage of multiple " Ioculations, " and eventual laser fenestration of the multiple "septa" to achieve adequate decompression of the hydrocephalus [2,5].

Clinical experience has shown: (1) that mental retardation in hydrocephalic children is directly related to the development of shunt infection [11 , 12], (2) that the most severe infections occur with Gram-negative organisms [2 ,13-15] , (3) that younger children with immature brains suffer the most severe sequelae , and (4) that the so-called multilocular ventricles are refractory to routine shunt decompression [1 , 2,5-9,15.] The following study was undertaken to increase our understanding of the pathophysiology of Gram­negative bacterial infection in the immature nervous system and to

serve as a guide to improved management of the infant with shunt infection.

Rationale for the Experimental Model

The hy-3 mouse is an inbred strain wi th a significantly depressed immune system and a genetic form of hydrocephalus transmitted as a mendelian recessive trait [15-'19]. Litters of hy-3 mice thu s provide hydrocephalic study subjects with otherwise similar but nonhydrocephalic litter-mates for controls. Because hydrocephalus in these mice is genetically determined, the model is free from artifact introduced by a destructive procedure undertaken to create hydrocephalus. Prior detailed descriptions of the histological and ultrastructural features of the normal and hydrocephalic hy-3 mouse brain provide an adequate basis for distinguishing between patho­logical changes caused by ventriculitis and those associated with hydrocephalus alone [15-19]. Prior work has documented that th e extracellular space of the hy-3 mouse does not reach full maturity and hydrocephalus is not fully expressed until age 21 days [17].

Materials and Methods

In the protocol designed and implemented by D. G. M . and Y. Y., the study group of 133 hy-3 mice was divided into three age groups simulat ing different degrees of brain maturity: 5-day-old mice (n =

39), 10-day-old mice (n = 4 8 ), and 15-day-old mice (n = 4 6 ). Of this group, seven 10-day-old mice and three 15-day-old mice were hydrocephalic. Twelve normal and six hydrocephalic hy-3 mice of various ages underwent insertion of sterile catheters to serve as controls. A virulent K 1 strain of E. coli was isolated from a docu­mented human case of shunt infection and maintained by serial transfers in our laboratory [20] . Inocula were prepared from well shaken broth suspensions diluted in normal saline to concentrations of 103-107 microorganisms/ ml.

A 20 gauge trocar and polyethylene tubing (outer diameter 0.61 mm) were inserted under sterile conditions through skull and frontal lobe into one lateral ventricle; 5 /.d of inoculum or control solution was injected , and the trocar was removed. The tubing was then cut and left in situ, covered by skin , to simulate a contaminated shunt system.

Animals surviving for 5 days were sacrifi ced and studied by light microscopy, elec tron microscopy, and photomicrog raphy of th e

This work was supported by the Laboratory for Oculo-Cerebrospinal Investigation (LOCI) , an arm of the Division of Ped iatri c Neurosurgery of Ch ildren 's Memorial Hospital and Northwestern University Medica l School.

1 Department of Radiology, Children 's Memorial Hospital and Northwestern Unive rsity Medica l School, 2300 Children 'S Plaza, Ch icago, IL 60614 . Address reprint requests to T. P. Naidich.

2 Division of Pediatric Neurosurgery, Chi ldren 's Memorial Hosp ital and Northwestern Unive rsity Medical School, Chicago, IL 60614 .

AJNR 4:461-465, May/ June 1983 0195-6108/ 83/ 0403-0461 $00.00 © American Roentgen Ray Society

462 CT OF THE HEAD AJNR:4, May / June 1983

Fig. 1.-Primary E. coli pyocephalus in hydrocephalic hy-3 mouse (H and E stain). Leukocytes pack inoculated ventric le (R) , surround choroid plexus (arrowheads ), and infiltrate its stroma. Choroidal epithelium and ependymal layer (arrows) are we ll preserved. There is little subependymal edema.

meninges, subarachnoid spaces, ventricles, corpus callosum, cath­eter, catheter tract , and ventriculostomy site. Animals dying of infection within 5 days were used to assess infection rates and mortality. Animals in whom technical complications such as punc­ture-induced hemorrhage obscured the pathologic findings were exc luded from analysis. Additional details of the experimental pro­cedures are reported elsewhere [15 ; other unpublished data].

Results

None of 18 control animals contracted infection. The overall infec tion rate in the study group of 133 was 89.4%. Five of the 10 hydrocephalic mice fail ed to contract infection and survived. All five hyd rocephalic mice in whom infect ion was successfu lly introduced died . Lin ear reg ression analyses indicated independent, statist ica lly significant increases in mortality with increase in numbers of orga­nisms injected; increasing immaturity of the mouse; and presence of hydrocephalus.

Postmortem Findings

Strikingly consistent anatomic and histologic changes are de­scribed below. Ultrastructural changes will be discussed in a sep­arate report. The polyethylene tube punc tured the ipsilateral ventri­cle in all cases. In some animals, particularly those in the 5-day-old group, the tip of the tube lay only partly within the ventricle, suggesting that inoculation was partly intraventricular and partly ex traventricular.

An abscess was usually present at th e shunt tip. Depending on the location of the tip, the abscess was a focal paraventricular abscess and / or a localized primary pyocephalus. When the inocu­lum produced primary pyocephalus, the infection remained largely confined to the ventricle and did not transg ress ependyma to infect the paraventricular region (fig . 1). In such cases, the ependymal layer appeared intact on light microscopy and the subependymal zone showed little evidence of edema or inflammation.

When the inoculum produced a paraventricular abscess, th e abscess frequen tly spread along the extraluminal surface of the ventricular wall and extended into the choroid plexus via the tela choroidea. The ependymal layer appeared intact on light micros­copy. In the main , the infect ion did not transgress the intact epen­dyma; only limited amounts of purulent material could be observed in the ventric le in such cases (figs. 2-4). Paraventric ular abscesses

Fig . 2.-Paraventricular inflammation in hydrocephalic hy-3 mouse with E. coli ventriculitis (H and E stain). Coronal section through inoculated hemisphere (R) , contralateral hemisphere (L) , and pus-filled interhemispheric fissure (arrowhead). Hemorrhage, edema, and inflammatory cells penetrate co rpus callosum to contralateral paraventricular reg ion. Contralateral ventri­c le shows little pus, normal choroid plexus, intact ependymal layer, and moderate edema of subependyma.

Fig. 3. -Paraventricular inflammation in hydrocephalic hy-3 mouse with E. coli ventriculitis (H and E sta in). Marked inflammatory edema has expanded extracellular space and penetrated corpus callosum from inoculated side (not seen) to contralateral paraventricu lar reg ion (L). Edematous spaces appear as small cysts (arrow) or larger coalescent cavities (arrowheads) filled with homogeneously staining fluid. Cavities vary widely in size.

frequently compressed the subjacent ventricle and occasionally isolated the lateral ventric les by occluding the foramina of Monro, causing substantial ventricular dilatation . Paraventricular abscess extended retrograde along the catheter tract in some mice, but the degree of infection was usually milder than at the shunt tip. In the younger an imals, particularly the 5-day-old group, the infection spread diffusely throughout the parenchyma and was often asso­ciated with focal necrosis in the paraventricular tissue.

Spread of infection in either direction across the ependyma was observed in three situations: (1) when the ependymal lining had been disrupted mechanically by the trocar; (2) when the infected mice were hydrocephalic (even if the infection at the primary site was mild); and (3) when the infected mice were very immature. When paraventricular infection transgressed the ependyma to cause secondary pyocephalus , the inflammatory celis generally remained attached to the ventricular wall locally and to the choroid plexus. They filled the ventricular cavity only in severe cases.

AJNR:4, May/June 1983 CT OF THE HEAD 463

Infection commonly spread to th e corti ca l subarachnoid space via either the CSF pathways or (less frequently) the catheter trac t. Subarachnoid accumulations of purulent materi al were most prom­inent in the basal cisterns. In severely infected animals, parti cularly those in the 5-day-old group , small subcorti ca l abscesses appeared to develop by direct extension of subarachnoid infections along the Virchow-Robin spaces.

Choroid Plexus

The choroid plexus was often totally surrounded by purulent material (fig. 1). The inflammatory ce lls coated the choroid plexus and extended into the stroma. The choroidal cell s rested on an intact basal lamina. However, the underly ing connective ti ssue

Fig. 4. -Paraventricular pseudoventri c les with " multiloculation" in hydro­cephalic hy-3 mouse with E. coli ventri culi t is (H and E stain). Coronal section through cortex and whi te matter of uninoculated contralateral hemisphere (L) shows uninoculated ventricle (V) ; equally large, juxtaventricular edematous cavity (C); and several small subependymal cysts (arrowhead). Pseudoven­tri cular edematous cavity (C) and small cysts shear ependyma (arrow) from brain and displace it inward. True ventri cle (V) and abutting pseudoventri cle (C) may be misinterpreted as single ventric le loculated by intravent ricular septations. Inset: High-power mag nification of junct ional reg ion. Ventricu lar wall is fo rm ed by smooth , intact ciliated ependyma (arrow), while equally smooth cyst wa ll is formed by compacted white matter (c/ose.d arrowhead) denuded of ependyma. " Septati ons" may be formed by white matter alone (open arrowhead) or by intact ependyma plus res idual subependymal white matter (arrow ).

Fig. 5. -E. coli ventriculitis in hydro­cephalic 3-year-old boy. A , Unenhanced axial CT. Appearance of loculati on and septation is simulated by large, noncom­municating callosal and juxtaventricular white-matter cyst (c losed arrowhead) that compresses lateral ventricles and displaces intraventricular shunt cathe­ters inferolaterally. Additional white-mat­ter cysts in parietooccipital reg ion (open arrowhead). B, Intraoperative photo­graph of apparent intraventricular sep­tum at point indicated by c losed arrOw­head in A. Laser fenestration of this thin , translucent outer wall afforded ex it of cyst fluid . Inset: Electron micrograph of "septum." Intact axons (arrowhead) show " septum " is residual whi te matter, not inflammatory membrane or glial scar.

A

stroma and fenestrated capillary network were involved in the inflammatory process.

Extracellu lar Space and Sub ependymal Region

The extracellular space was markedly increased in many animals, espec ially in the subependymal region. This increase was more dramatic in the hydrocephalic animals and in the youngest animals,

particularly the 5-day-old group. In most animals, inflammatory ce lls were concentrated in th e subependymal reg ion despite the absence of identifiable organisms in th is reg ion. On ly in the very immature 5-day-old group were organisms consistently observed. Infl ammatory ce lls and edema consistentl y spread from the paraventricular region to the contralateral hemisphere along the enlarged ext racellular spaces of the corpus callosum and other commissures, spread ing to d istan t sites of the inoculated hemisphere along th e enlarged extracellul ar spaces of the longitudinal nerve bund les (fig . 2). In­fl ammation nearl y always reached the opposite paraventricular reg ion more rapidly via the extracellular spaces of these fiber bundles than by the CSF pathways of the ventric les (figs. 2 and 3). Inflammation also involved the extreme capsules prominently.

The enlarged, edema-filled extracellul ar spaces appeared to form small caviti es or cysts in the white matter, parti cularly in the para­ventricular reg ion (fig. 4). These cysts separated the ce llular ele­ments of th e white matter and, in places, caused the in tact epen­dymal layer to detach from the brain parenchyma (fi g. 4) . The smaller cysts appeared to coalesce into larger cysts, producing some edema cavities as large as the ven tric les themselves. Such cavities may be deSignated pseudoventri c les. When two large edema cavities abutted each other , the intervening residual whi te matter had the appearance of a thin septum (fig. 4). When a juxta ventricular edema cavity compressed the ventric le and dis­

placed the intact, detached ependyma inward , the combined cavi­ties of the ventric le and the pseudoventric le resembled a sing le dilated ventric le loculated by a thi n in traventr icular septum or ve il (fig. 4). Where abu tting cyst walls began to diverge from the ventricle or from each other , the remaining brain tissue typ ica lly assumed the shape of a small tri ang le or wedge whose apex gave ri se to the septum (fi g. 4).

Neuropil

The neuronal elements and myelin of the brain appeared relati ve ly resistant to the inflammatory process. Glial astrocytic processes,

8

464 CT OF THE HEAD AJNR:4, May / June 1983

however , appeared very sensitive to the infection. Th e peri vascular glial foot processes appeared retracted from the endothelial cells, leaving definite c lear spaces between. Large areas of edema were almost devoid of astrocytic processes, leaving neurons and synaptic complexes devoid of glial investment. This shrinkage and loss of the glial component of the brain was a dramatic feature of the pathology of ventriculitis.

Discussion

In mice and man, ventr iculitis causes the greatest damage to the brain in younger animals, hyd rocephalic animals, and those with disruption of the ependymal layer. One anatomic feature common to both young animals and hydrocephalic animals is an unusually large extracellular space [1 5-1 9]. Our study shows that severe ventriculitis is associated with substantial ex tracellular edema and with spread of inflammation along th e dilated extracellular spaces of th e commissural and longitudinal fiber tracts to distant sites in the inoculated and contralateral hemispheres. If the extracellular space is already large at the time of inoculation, as in immature and in hydrocephalic animals, then th e spread of infection is likely to be more rapid and the infection more severe. In animals with depressed immune response, such spread may overwhelm the animal before its immune system is able to respond. These factors probably explain the greater severity of ventriculitis in hydrocephalic an imals and the earl y deaths of all infected hydrocephalic mice . They may also shed light on the greater severity of Gram-negative ventriculitis in very young human infants.

In this and previous stud ies [1 0], the intact ependyma seemed to be a significant barrier to the spread of infection into or out of the ventricle. The greater severity of ventriculitis and paraventricu lar infection in the hydrocephalic animal may also be related to the separation of ependymal ce lls and the deterioration in the ependy­mal " barrier " that occur with hydrocephalus. By all three ri sk factors, premature infants with hydrocephalus, prior disru ption of the ependyma by rupture of subependymal hemorrhage into the ventricle, and later disruption of the ependyma by shunting would be expected to suffer the most severe sequelae of shunt infection and ventriculitis. In our c linical experience, this has proved true.

The hy-3 mice wi th sterile hydrocephalus show little change in the corpus callosum [1 6-1 9]. The dramatic spread of inflammatory edema along the corpus callosum and th e formation of coalescent callosal and paraventricular edematous cavities appear to be spe­cific features of ventriculitis in these animals. The marked reduction in the glial astrocytic processes also appears to be specific to ve ntr icu liti s. Loss of the astrocytic processes probably reduces the structural integrity of the white matter and contributes to th e for­mation of the large edematous cavities.

In our opinion, processes ak in to those demonstrated in this murine model may explain some instances , perhaps most instances, of the so-called mult iloculat ion and septation of the ventr ic les in human Gram-negat ive ventriculitis (fi g . 5A) . Certainly th e coalescent edematous cavities may grow large, may compress the ventr ic les, may abut each other (thus leaving thin walls of in tervening white matter), may abut and compress the ventri c les, and may displace the intact ependymal layer inward as an intervening sheet of ti ssue (fig. 4). Th e edematous cavities appear smooth-walled although they are not lined by ependyma, and they are f illed with fl uid of a density very sim ilar to that of the stagnant CSF. The wedges of t issue remaining between abutt ing cavi ties bear strik ing resem­blance to the glial tufts described by Schultz and Leeds [7]. These authors' descriptions of in flammatory cells; subependymal gliosis; small areas of denuded ependyma; glial tu fts, which extended through the denuded ependyma into the ventricular lumen; and

apparent origin of filmy, translucent fibroglial membranes from the glial tufts [7] might well be applied to the changes observed in the hy-3 mouse (figs. 1-4). Documentation th at axons traverse such membranes in human E. coli ventriculitis (fig . 5B) is strong support for the theory we advance. The common locations of the noncom­municating cysts in the corpus ca llosum and paraventricular re­gions; the failure to drain these spaces by intraventricular catheters; and the common observation that intraventricular catheters are usually displaced centrally by the undrained spaces also favor our theory.

Gilles et al. [21] have documented th at injection of sterile E. coli endotoxin into the peritoneal cavities of neonatal kittens (but not adult cats) produces severe paraventricular leukoencephalopathy and cavitation . In these kittens, a rim of ti ssue is preserved between the cyst and the ventric le and resembles a septum in some cases [21]. Th ese findings could help explain the increased inc idence of " multiloculation" and "septation" in patients with Gram-negative ventriculitis , especially E. coli ventriculitis.

We acknowledge that inflammation may scar the ventricular foramina to cause hydrocephalus with multiple , isolated ventricular chambers. This is not the process we discuss. True intraventricular septa and consequent ventricular loculation may occur in some cases. However, we have not observed these to date. The experi­mental data from the hy-3 mouse model of E. coli ventriculitis do not support the concept that a subependymal gliotic process dis­rupts the ependyma and produces glial tufts that act as niduses fo r the organization of intraventricular exudate and debris into filmy fibrog lial membranes. In our experience, it is difficult to grow glial cells even under optimal laboratory conditions. It is correspondingly unli kely th at such glial or fibrous ti ssue would grow down through a pus-filled cavity to form a septum across the ventri c le. Rather, the experimental and cl inical data are consistent with the conclusions that: (1) so-called multiloculation represents compression of the true ventricles by multiple , juxtaventricular white-matter cysts, and (2) the septa themselves represent displaced ependyma and resid­ual sheets of white matter traversed by axons. If this is true, then wide fenestration and disruption of the septa may be contraindi­cated , lest they disrupt functional pathways.

REFERENCES

1. Salmon JH. Isolated unilateral hydrocephalus following ventric­uloatrial shunt . J Neurosurg 1970;32: 219- 226

2. Kalsbeck JE, DeSousa AL, Kleiman MB, Goodman JM , Fran­ken EA. Compartmentalization of the cerebral ventr ic les as a sequela of neonatal meningitis. J Neurosurg 1980;52: 547-552

3. Gilles FH , Jammes JL, Berenberg W. Neonatal meningitis. The ventri c le as a bacterial reservoir. Arch Neuro l 1977;34: 560-562

4. Hill A, Shackelford GO, Volpe JJ. Brief cl in ical and laboratory observations . Ventriculitis with neonatal bacterial men ingitis: identification by real-time ultrasound. J Pedia tr 1981 ;99: 133-136

5. Rhoton AL Jr, Gomez MR. Conversion of multilocular hydro­cephalus to unilocu lar. Case report . J Neurosurg 1972;36: 348- 350

6. Albanese V, Tomasello F, Sampaolo S. Multiloculated hydro­cephalus in infants. Neurosurgery 1981 ;8: 641 - 646

7. Schultz P, Leeds NE. Intraventricular septations complicating neonatal men ingitis. J Neurosurg 1973;38: 620- 626

8. Savolaine ER , Gerber AM . Computerized tomog raphy stud ies of congenital and acqu ired cerebral intraventricu lar mem­branes. J Neurosurg 1981 ;54: 388- 39 1

AJNR:4, May / June 1983 CT OF THE HEAD 465

9. Brown lW, Zimmerman RA, Bilaniuk LT. Polycystic brain dis­ease complicating neonatal meningitis: documentation of evo­lution by computed tomography. J Pediatr 1979;94: 757 -759

10. Berman PH , Banker BQ. Neonatal meningitis. A c linica l and pathologica l stud y of 29 cases. Pedia trics 1966;38: 6- 24

11. Brown JT, Mc l one DG. The effec t of complications on intellec­tual function in 167 children with myelomeningocele. Z Kin­derchir 1981 ;2: 117-1 2 1

12. Mcl one DG , Czyzewski 0 , Raimond i AJ , Sommers RC. Central nervous system infections as a limiting factor in th e intelligence of children with myelomeningocele. Pediatr ics 1982;70 : 338-342

13. Lorber J , Pickering D. Inc idence and treatment of post-men­ingitic hydrocephalus in th e newborn. Arch Dis Child 1966;41 :44 - 50

14. Smith ES. Purulent meningitis in infants and children. A review of 409 cases . J Pediatr 1954;45 :4 25-436

15. Mc l one DG , Killion M, Yog ev R, Sommers MW. Ventriculiti s of mice and men. In: American Society for Pediatric Neurosur-

gery, ed . Concepts in pediatric neurosurgery, vol. 2. Basel: Karger, 1982 : 11 2-1 26

16. Mc l on e DG, Bondareff W, Raimondi AJ. Brain edema in the hyd rocephalic hy-3 mouse. J Neuropathol Exp Neurol 1971 ;30 : 627 -637

17. Mc l one DG, Bondareff W, Raimond i AJ . Hydrocephalus hy-3 a murine mutant: II. Changes in the extracellular space. Surg Neuro / 1973 ; 1 : 233- 242

18. Mc l one DG. CSF pathways in a murine mutant. Childs Brain 1979;5 : 292- 303

19 . Raimondi AJ , Bailey 0 , Mc l one DG, Echeverry A. The patho­physiology and morphology of murine hydrocephalus in hy-3

and ch mutants. Surg Neurol 1973; 1 : 50-55 20 . McCracken GH Jr, Sarff lD, Glode MP, et al. Relation between

Escherichia coli K1 capsular polysaccharide antigen and c lin­ical outcome in neonatal meningitis. Lancet 1974;2: 246-250

21. Gilles FH , Averill DR Jr, Kerr CS. Neonatal endotox in enceph­alopathy. Ann Neuro/ 1977 ;2 :49-56